International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 3, June 2022, pp. 3309~3319
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i3.pp3309-3319  3309
Journal homepage: http://ijece.iaescore.com
A data quarantine model to secure data in edge computing
Poornima Mahadevappa, Raja Kumar Murugesan
School of Computer Science and Engineering, Taylor's University, Selangor, Malaysia
Article Info ABSTRACT
Article history:
Received Jun 4, 2021
Revised Jan 18, 2022
Accepted Jan 31, 2022
Edge computing provides an agile data processing platform for latency-
sensitive and communication-intensive applications through a decentralized
cloud and geographically distributed edge nodes. Gaining centralized control
over the edge nodes can be challenging due to security issues and threats.
Among several security issues, data integrity attacks can lead to inconsistent
data and intrude edge data analytics. Further intensification of the attack
makes it challenging to mitigate and identify the root cause. Therefore, this
paper proposes a new concept of data quarantine model to mitigate data
integrity attacks by quarantining intruders. The efficient security solutions in
cloud, ad-hoc networks, and computer systems using quarantine have
motivated adopting it in edge computing. The data acquisition edge nodes
identify the intruders and quarantine all the suspected devices through
dimensionality reduction. During quarantine, the proposed concept builds the
reputation scores to determine the falsely identified legitimate devices and
sanitize their affected data to regain data integrity. As a preliminary
investigation, this work identifies an appropriate machine learning method,
linear discriminant analysis (LDA), for dimensionality reduction. The LDA
results in 72.83% quarantine accuracy and 0.9 seconds training time, which is
efficient than other state-of-the-art methods. In future, this would be
implemented and validated with ground truth data.
Keywords:
Data analysis
Data integrity
Data security
Edge computing
Quarantine
This is an open access article under the CC BY-SA license.
Corresponding Author:
Raja Kumar Murugesan
School of Computer Science and Engineering, Taylor's University
Jalan Taylors, 47500 Subang Jaya, Selangor, Malaysia
Email: rajakumar.murugesan@taylors.edu.my
1. INTRODUCTION
Edge computing is a distributed computing paradigm that brings computation and storage closer to
the proximity of edge devices. These benefits have fueled many use cases like artificial intelligence (AI),
robotics, machine learning and Telco network communications and solved key challenges like bandwidth,
latency, resilience, and data sovereignty. The motive of edge computing is to provide a decentralized cloud
with low latency computation, overcome resource limitations of edge devices, and deal with network traffic
and data explosion [1]. Figure 1 shows an edge computing framework that includes edge devices, edge servers
and gateways as the essential components of the edge computing layer. Any device with computing, network
and storage capability can act as an edge device. They gather data from the internet of things (IoT) devices,
perform real-time data analysis, and respond faster than cloud computing. The gateways are responsible for
translation service between various heterogeneous devices in the edge and IoT and cloud layers. The edge
servers manage several edge devices and gateway by handling the context data. It is a generic term that captures
associated computing paradigms such as fog computing, cloudlet, or mobile access edge computing [2].
Although edge computing provides many benefits, real-time data analytics that relies on edge
computing faces various security and privacy issues. Edge data analytics includes data collection from different

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 3309-3319
3310
IoT devices, storing, processing, and analyzing them on geographically distributed edge nodes. The data
owners lose control over the data transmitted to the edge nodes, and even achieving centralized control over
them can be challenging. During this, the edge nodes may become vulnerable to intruders who can control the
network, compromise the edge nodes, alter, or modify the transmitted data. These actions of the intruders can
create integrity issues that result significantly in degrading the efficiency of edge data analytics and the
performance of edge-based applications [3]. Therefore, an efficient data security solution in edge computing is
required to ensure data integrity during edge data analysis and retain the performance of edge-based
applications. Currently, there are many data integrity verification solutions to secure the data in edge
computing, such as using blockchain devices, data assessments, collaborative methods or statistical anomaly
detections [4]. These solutions increase resources utilization of lightweight edge nodes, adds computational
load, or develop frequent interaction with intruders to understand their behaviors. Hence, the proposed research
intends to address these issues and develop a lightweight security solution that does not drain the resource-
limited edge nodes.
Figure 1. Edge computing framework with the dataflow
The proposed paper addresses data integrity issues in edge computing by identifying any intrusions
caused in the network and later quarantine all the suspected devices and their data. The combined intrusion
detection and data quarantine model identify the legitimate devices through the reputation score. Further, the
confirmed legitimated devices data is sanitized based on the scrubbing score to recover the original data
affected due to intrusion and thereby regain data integrity. The cyber security systems have used the same
concept to quarantine the infected nodes from the computer system till their recovery [5]. In the cloud
computing system, real-time data security using quarantine methods for petabytes of data from viruses and
Trojans is successful. In addition, there are many patent models for data security in file systems, databases, and
computer systems. Due to these reasons, the proposed work adopts a data quarantine model to address the data
security issues in edge computing. It is important to note that the quarantine approach is a renewed concept in
edge computing. The proposed method does not add any computational load to the resource-constrained edge
nodes, create alarms or allow suspected devices to interact with the other edge resources.

Int J Elec & Comp Eng ISSN: 2088-8708 
A data quarantine model to secure data in edge computing (Poornima Mahadevappa)
3311
In the proposed concept, the quarantined devices will not be aware of the isolation and transmit data
continuously. The spam detection module here uses this transmitted data to analyze their behavior and identify
legitimate devices. The data scrubber module sanitizes the data of these legitimate devices and sends it to the
edge nodes. The sanitized data recovers the original data that was affected due to intrusion and assist in
regaining the data integrity. As a preliminary work for the proposed concept, this paper identifies an appropriate
machine learning (ML) for intrusion detection and quarantine. Overall, the objectives of the proposed concept
are identified: i) to identify the attack and isolate the infected data efficiently without alarming the nearby edge
nodes, ii) to prevent the illegitimate users from sending data to the edge nodes, and iii) to cleanse the inaccurate
data and formulate sanitized data for edge data analytics.
The decentralized edge computing paradigm complements cloud computing by optimizing the
performance of user-driven and communication-intensive applications. These features have fascinated many
IoT applications to adopt edge computing and benefit from it. Smart fog hub services (SFHS) is an edge-to-
cloud platform deployed in Cagliari Airport in 2019 to provide a user-friendly and indoor map to the passengers
to get their way to preferred shops, restaurants or any services during their waiting time in the airport [6]. This
application promises the best customer experience and increases revenue for the airport and service providers.
They use a recommender engine to track the users to infer their tastes and preferences based on their actions
or similarities [7]. Many airports like Singapore provide smart airport services like biometric passports, Radio
Frequency Identification (RFID) to track baggage, use beacons to track passengers, and many more. There are
around 28% of smart airports in the world to augment users experience [8]. But this experience may create new
cybersecurity challenges. There are many incidences wherein September 2018, British Airways was fined
£183Million for a data breach in a security system that affected 380,000 transactions [9]. In May 2018, a data
hack in Iran's Mashhad airport displayed protest messages in airport monitor's [8]. The hacking was mainly
due to a privilege escalation attack. Therefore, when adopting these technologies and storing or sharing the
data at the user's proximity, it is necessary to adopt security measures to safeguard data confidentiality and
integrity. For instance, these data hacks by spammers or attackers can introduce undesirable data and strain
into the system.
The rest of the paper is organized: section 2 provides an overview of edge data analytics and data
security issues in edge computing. Section 3 discusses the existing literature review. Section 4 presents the
research methodology of the proposed concept. Section 5 evaluates the preliminary results. Section 6 discusses
the obtained results, and finally, a conclusion with future work is included.
2. EDGE COMPUTING AND DATA SECURITY ISSUES IN EDGE LAYER
The standard computing framework, FogFlow developed by the NEC laboratory, is used as a ground
framework, and deploy the proposed concept. This framework supports standard interfaces to share and reuse
contextual data across services. As shown in Figure 1, there are three logical divisions in the framework: service
management, data processing, and context management. The service management includes topology master
(TM), task designer and docker image repository. Task designer provides a web interface to monitor the IoT
services, and the docker repository manages all the docker images. TM is responsible for service orchestration
to handle service requirements and service topology among the edge nodes. The worker or edge nodes at the
proximity of IoT devices perform data processing tasks assigned by TM. TM and workers communicate
through RabbitMQ protocol. Finally, context management includes IoT discovery, a set of IoT brokers and
federated brokers. These components establish data flow across the tasks and manage contextual data like
availability of workers, topology, task and generated data stream [10]. Figure 1 also depicts the process and
data flow of the framework. The FogFlow framework supports various use cases like lost child finder, anomaly
detection in smart cities, smart parking, and smart industry. Therefore, implementing the proposed data
quarantine model (DQM) considering the FogFlow framework can support the real-time applicability of the
concept to address data integrity issues in any use case scenarios.
2.1. Data security issues
Edge computing handles data from various sensors, devices, servers, including local data centers and
centralized cloud. The edge nodes gather data from the different ubiquitous devices, transmit it to the other
edge nodes or servers, and analyze it. The data analysis includes task deployment defined as a mobile agent
and deployed dynamically on the edge nodes. These tasks can be parallel, asynchronous and sometimes
independent of one another, without the intervention of the other nodes [11]. The data collection, transmission,
and task deployment process provide appropriate real-time interaction and monitoring of the applications,
known as edge data analytics. However, decentralized edge data analytics are vulnerable to security threats due
to a lack of centralized control and the features of edge nodes such as heterogeneousness, mobility, and
geographical distribution. The vulnerabilities can create various attacks that affect data confidentiality,
integrity, authentication, and access control. Figure 2 shows various threats that cause these data security issues.

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 3309-3319
3312
Figure 2. Category of data attack in edge computing
2.1.1. Data confidentiality
It is a fundamental requirement in the edge layer to ensure that the data owners and users can access
private information without the intervention of unauthorized users. The edge nodes receive data in the core
network infrastructure, transmit and process it to the other edge nodes or cloud data centers [12]. The attackers
can capture the data packets during communication and read sensitive information like passwords, usernames,
and credit card details during this process. If one or more edge nodes are affected, the adversaries can wiretap
the entire system and gradually decode the data packets [13]. Therefore, the security frameworks must manage
and store data without compromising the confidentiality in the edge layer. Common approaches used to achieve
data confidentiality are cryptography schemes, secure data acquisition and secure IoT devices or edge nodes.
Dynamic keys encrypt, authenticate, and fragment input data in cryptography schemes. In a secure data
acquisition, retransmission of decisions for each communication optimizes the data. Apart from this,
Blockchain or Artificial Intelligence (AI) are incorporated on IoT devices or edge nodes to validate the data
periodically.
2.1.2. Data integrity
This is a measure to refer originality, completeness, and accuracy of the data from the sourcing until
the entire data analysis process. The data owners lose control over their data while outsourcing to the edge
nodes, and thereby the data can be vulnerable to any security attacks. The attacks can modify, alter, or delete
the data with malicious intent creating integrity issues [14]. Lack of data integrity and correctness solution can
affect the performance and efficiency of edge computing. The possible solutions to achieve data integrity are
batch auditing or dynamic auditing using homomorphic tags to the outsourced data. There are few privacy-
preserving protocols to enhance security through monitoring the system.
2.1.3. Data authentication
It is the process to validate the identity of users and guarantee that the users are legitimate to access
cloud servers or edge nodes. In addition, it is necessary to authenticate the participating edge nodes and edge
data centers. By establishing authentication in edge computing, data quality can be improved drastically [15].
Password, smart card, biometrics-based authentication are the typical authentication schemes adopted in cloud
and edge computing applications.
2.1.4. Accessibility
This is a mechanism for determining the rights and privileges of edge nodes and users in the system.
All the users in the network cannot access sensitive information. Hence, it is necessary to specify local policies
to determine access specifications for users and resources in the infrastructure. In edge computing, choosing
accessibility can be for three reasons: storage and computation services, edge nodes to access particular
resources, and virtual machines [16]. Some of the access control mechanisms adopted in edge computing are
access control models like–mandatory access control, discretionary access control, role-based access control
and Attribute-based access control.

Int J Elec & Comp Eng ISSN: 2088-8708 
A data quarantine model to secure data in edge computing (Poornima Mahadevappa)
3313
Among all the above data attacks, the data integrity issues severely impact the data stored and
processed in edge computing. The issues can lead to inconsistent and unreliable information leading to the
wrong decision during edge data analytics. Therefore, the proposed method ensures integrity issues caused due
to any intrusion resulting in data modification, alteration, or injection to the transmitted data in edge computing
and present a concept to handle these issues. False data injection, structured query language (SQL) or route
injection, spam, data falsification, malware, replay are some threats that cause data integrity issues through
data fabrication or modification [17]. It is necessary to address these issues immediately since their impact can
propagate from one edge node to another and reduce the performance by affecting the data analysis process. In
addition, it would be challenging to identify the root cause and results in additional repair costs and delay
recovery [18]. Hence, we propose an analytical method to detect integrity attacks due to false data injection by
employing a data quarantine model.
3. LITERATURE REVIEW
Data integrity attack or integrity attack is an attempt to corrupt the data intentionally by modifying,
deleting, or fabricating during data outsourcing. Data, route or SQL injection, malware, email spoofing, replay
are the possible attack vectors to affect data integrity. Through this attack, the attackers can understand the
communication protocol and gain control over the system resulting in unstable operations or prolonged loss of
resources. The impact of the attack is proportional to the duration of the attack [19]. There are limited works
in edge computing to address data integrity issues, and Table 1, summarizes these related works. Blockchain
is a decentralized ledger that manages multiple participants collaboratively without centralized control. Both
edge computing and blockchain have distributed architecture; hence adopting it here can help achieve
significant interdependency. In distributed neural networks, blockchain tracks end-to-end IoT based
applications. An efficient inter-blockchain querying and locking mechanism in the edge layer improves
security and performance [20].
Similarly, blockchain is adopted to ensure data integrity during the task offloading process. They
employ simple additive weighing and multi-criteria decision-making techniques to confirm the optimal
migration of virtual machines. But achieving optimal migration requires many iterations, which is time-
consuming [21]. Despite the advantage of blockchain providing centralized monitoring of the application, they
are resource hungry and immutable for stored data. These features of blockchain can be a great threat to edge
computing applications; however, further research in this area can solve many issues.
Table 1. Related work
Reference Working model Attacks addressed Limitation
[20] Blockchain-enabled integrity protection
scheme in distributed neural networks
Data poisoning and
model poisoning
Blockchain increases resource utilization
of lightweight edge nodes
[21] Blockchain-enabled integrity preserving
system during task offloading
Unauthorized transaction Blockchain increases resource utilization
and makes stored data immutable
[22] Collaborative bad data detection scheme in
the electric metering system
False data injection
destroying data integrity
and availability
Increases computational load to edge
nodes
[23] Damage assessment and data recovery to
ensure consistent database in smart city
applications
Malicious activity and
privacy violation
Increases computational load to edge
nodes
[24] A statistical learning-based anomaly
detection system in smart grid application
Data integrity attack due
to false data injection
Additional computational load to track
disseminated data is included
A collaborative bad data detection scheme identifies integrity issues due to false data injection in an
electric metering system. It includes a set of rules on the edge nodes to determine the reputed electric meter.
These rules identify false data injection that destroys the integrity and availability of the data [22]. Damage
assessment and data recovery system for smart city applications is another solution to recover data from
malicious attacks. This solution recovers original data and returns the database to a consistent state for data
analysis in edge computing. This approach has three recovery algorithms: main damage assessment, secondary
damage assessment, and main recovery for each data transaction. This process continues till all the affected
transactions in the entire system are detected [23]. This approach includes detection schemes on each edge
nodes and increase the computational time of lightweight edge nodes.
Anomaly detection for data integrity attack in smart city application is used to secure data and privacy
in the edge layer. Edge nodes in the higher hierarchical layer that has access to inter-area data have anomaly
detectors. They regularly scan the input data stream to ensure data integrity in the entire system [24]. Although
this approach provides a conventional centralized approach to obtain results with reliable decisions,

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 3309-3319
3314
guaranteeing the data integrity can be a tedious process. Edge nodes at the lower level would have disseminated
the data to other edge nodes, so tracking the data will add computation to the edge nodes. Overall, the discussed
security solutions have not considered mitigating the attack early during data acquisition. Addressing the issue
during data acquisition can assist in data tracking and identify the root cause of the issue.
4. RESEARCH METHOD
This paper proposes a data quarantine model to defend against data integrity attacks. The proposed
model considers continuous measures to quarantine the affected data without affecting data analysis in the edge
layer. Figure 1 shows intrusion detection includes in the existing FogFlow framework through feature
extraction. The worker nodes fetch the data from the IoT devices, analyze data and identify any intrusions.
After identifying the intrusions on the edge nodes, the edge server-IoT broker quarantines the infected data for
a predetermined time in DQM. Figure 3 shows the modules in the DQM, and the following section includes
the description of each module.
Figure 3. Proposed data quarantine model
4.1. Short time quarantine data storage
Short time quarantine data storage is a temporary database to store intruded devices and their data.
The storage includes device ID, device location, worker ID, and the sourced data. This data is passed to the
spam detection module to identify spam and generate the devices reputation and spam scores. This data storage
stores these scores that assist in predicting the behavior of the devices. Later passes it to the data scrubbing
module to perform data sanitization. This temporary data storage stores all these scores for a predetermined
time until the edge resources receive the sanitized data. The untrusted data and devices information is discarded
to restore the data storage.
4.2. Spam detection module
The spam detection module analyses the behavior of the devices based on the data they transmit during
quarantine. The spam classification matrix categorizes the data as spam or non-spam using and Table 2 shows
the spam confusion matrix. The confusion matrix includes the following values: TP–(true positive) represents
the spam correctly classified, FN–(false negative) spam misclassified as a spammer, FP–(false positive) Non-
spammer misclassified, and TN–(true negative) is the number of non-spammers classified correctly. The spam
score generator generates the spam score for the classified spam data using (1).

Int J Elec & Comp Eng ISSN: 2088-8708 
A data quarantine model to secure data in edge computing (Poornima Mahadevappa)
3315
Table 2. Confusion matrix
True class
Positive Negative
Predicted class
Positive TP FP
Negative FN TN
To obtain spam scores the as (1) is used
𝑆𝑆 =
(𝑁𝑆𝐷+(𝑁𝑀𝐷 𝑋 𝑁𝑆𝑆))
𝑆𝐼
(1)
where, SS–spam score, NSD–number of attacks sent to the same destination, NMD–number of attacks sent to
multiple destination, NSS–number of attacks sent by single attacker, and SI–spam interval
In (1), SI determines the time interval between the two attacks. If the time interval between two attacks
is small, the spam score is higher. If the attackers send the attack to the same or different destination, then NSD
and NMD increase. Similarly, if a single attacker sending attacks to many destinations, NSS is considered.
Based on this score, a SS of 9 is considered a threshold and used in reputation updater to identify legitimate
devices.
4.3. Reputation updater
Reputation updater and spam score is bi-directional feedback to each other. It includes a reputation
table with device id and reputation score. Based on the spam score, the reputation of each device is measured.
Figure 4 shows the relation between these scores in the spam detection module. If the devices have a spam
score greater than the threshold, they are blacklisted devices, and IoT brokers can prevent receiving data from
blocked devices. If the past data in the spam module has the least spam score, then the devices are considered
a whitelist and are authorized to transmit data to the edge layer.
Figure 4. Illustration of reputation score
4.4. Data scrubber
The data scrubber module sanitizes the inconsistent, uncertain, and ambiguous trusted data of
legitimated devices stored in a short time data storage. First, they prefetch the data from the spam classifier to
convert and consolidate it into a single format. Later several rules are applied to validate the consistency of the
data using scrubbing scores. Finally, the data is loaded back to the database and sent to the IoT broker for edge
data analytics. The sanitized data obtained improves the data quality, integrity and enhance the edge decision
support system.
5. RESULTS
The proposed work implemented intrusion detection on the worker edge nodes during data acquisition
as preliminary work. The worker nodes process the acquired data to obtain a high-degree spatial separation of

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 3309-3319
3316
high dimensional data. This dimensional reduction monitors the input data stream and updates any changes to
the IoT broker. The main objective of updating changes is to identify any attacks and quarantine the suspected
devices and their data. IoT broker decides to pass the data either to the quarantine model or the available edge
nodes for data analysis based on the signs of a possible attack. This whole approach of data quarantine and
indicating attacks in the devices do not add any load to the existing edge nodes or create notifications to other
resources.
5.1. Dimensionality reduction
Dimensionality reduction is a machine learning (ML) technique to reduce the higher dimensional data
to lower dimensions that remove redundant and dependent features. To identify the efficient ML method for
feature reduction, the following ML methods are analyzed, like linear discriminant analysis (LDA), logistic
regression (LR), and support vector machine (SVM) and nonlinear classifier multi-layer perceptron (MLP).
The comparative study considers the NSL-KDD dataset and evaluates these methods in terms of training time,
training accuracy and quarantine accuracy.
5.2. Experimental setup
The proposed work implemented the ML dimensionality reduction techniques on Python-based yet
another fog simulator (YAFS) [25] using the NSL-KDD dataset [26]. The same tool will be used to implement
the proposed model in future. The dataset consists of varying proportions of normal and attack data set. The
simulation setup includes a cloud node of 10 GB RAM and 16 GHz CPU, two edge nodes of 2 GB RAM and
3 GHz CPU, and 50-400 IoT devices of 500 MB RAM and 1 GHz CPU power. The links bandwidth is
3-10 Mbits, the packet size of 200 bytes and 100×108
packets per instructions.
Figure 5 shows the results obtained by the feature extractions using the methods mentioned above.
Figure 5(a) shows that LDA and LR has a minimum training time of 0.9 seconds and 1.5 seconds, respectively
since LDA is simple, efficient and has less computational overhead. The LDA class matrix of dimensionality
reduction is inverse of the grouped covariance matrix, which transposes input data and class mean vectors that
make it very simple. Similarly, in LR, a normalization technique is applied to achieve the objective using the
sigmoid function. Figure 5(b) shows the training accuracy of these methods and notice that MLP has 99.64%
accuracy, but the training time is 88.95seconds which is considerably more than LD and LR. MLP is a
particular case of supervised artificial neural networks (ANN). This network architecture includes an input
layer, one or more hidden layers with some number of neurons, and an output layer with one neuron for each
class to be classified. Therefore, it is possible to achieve higher accuracy by comparing each value of the
distribution vectors
The proposed quarantine aims to send infected nodes to the quarantine, and hence LDA and LR
techniques are further analyzed to achieve better precision towards devices and data quarantine. Quarantine is
forced isolation or stoppage of the interaction of the infected device and data with the edge nodes in the
framework. Figure 5(c) shows the quarantine accuracy of LDA and LR techniques, and it shows that LDA has
72.83% accuracy in sending the devices to the quarantine section.
5.3. Evaluating the devices and data sent to quarantine
As noted, LDA and LR has better quarantine accuracy compared to SVM and MLP. This section
analyses LDA and LR to identify efficient quarantine techniques. The following equation determines the
accuracy of these methods:
𝑇𝑃 + 𝑇𝑁
𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁
where
TP (true positive): number of instances correctly classified as an attack.
FP (false positive): number of instances wrongly classified as an attack.
TN (true negative): number of instances correctly classified as normal.
FN (false negative): number of instances wrongly classified as an attack.
Figures 6(a) and 6(b) show that the LDA quarantines 265 devices out of 400 and 2047 data packets,
respectively. Figure 6(c) shows that the quarantine accuracy of LR with 50 IoT devices is slightly lesser
compared with the higher number of IoT devices. But the accuracy of LDA is 72% irrespective of the number
of IoT devices. This quarantine accuracy shows that LDA is more efficient and reliable in quarantining the data
and IoT devices regardless of the number of devices. Therefore, using LDA techniques, the proposed DQM
will be implemented further and address the data integrity issues caused due to any intrusions in edge
computing.

Int J Elec & Comp Eng ISSN: 2088-8708 
A data quarantine model to secure data in edge computing (Poornima Mahadevappa)
3317
(a) (b)
(c)
Figure 5. Comparing the efficiency of the diffent machine learning algorithms (a) training time, (b) training
accuracy, and (c) quarantine accuracy
(a) (b)
(c)
Figure 6. Evaluating perfomance of the proposed qurantine model (a) devices quarantine, (b) data quarantine,
and (c) quarantine accuracy
0
50
100
150
200
250
300
50 100 150 200 250 300 350 400
Simulation
Time
Total Iterations
Number of Quarantined Devices
LDA LR
0
500
1000
1500
2000
2500
50 100 150 200 250 300 350 400
Simulation
Time
Total Iterations
Number of Quarantined Data
LDA LR
0%
20%
40%
60%
80%
100%
50 100 150 200 250 300 350 400
Accuracy%
Total Iterations
Quarantine Accuracy
LDA LR
Method Method
Method

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 3, June 2022: 3309-3319
3318
6. DISCUSSION
The preliminary results obtained in the proposed work includes the ML methods for intrusion
detection. The worker nodes in the FogFlow framework perform the dimensionality reduction of the data to
identify any intrusive activities. The training time of 0.9 s and 1.5 s of LR and LDA methods shows that these
methods identify intrusions faster. The faster detection would limit the attacker presence in edge resources and
do not provide any possibilities for the attackers to understand the behavior of the framework. Subsequently,
this can significantly reduce the intensity of the attack on the framework. Although the considered ML methods
have more than 90% accuracy, the proposed research focuses on quarantining the affected devices. Hence,
while analyzing the quarantine efficiency, LDA provides constant 72% accuracy regardless of the number of
IoT devices. Therefore, LDA was considered to have a better quarantine accuracy and efficient training time
of 0.9 seconds compared to other methods.
7. CONCLUSION AND FUTURE WORK
Edge computing is a promising architecture that provides a data processing platform between cloud
and edge nodes. Communication intensive and latency-sensitive applications have more significant benefits
due to edge computing. But security and privacy issues are challenging when the computation is brought closer
to the edge nodes. It can be observed from the literature that data attacks can compromise the edge nodes to
gain access into the network and affect data confidentiality, integrity, availability, and access control. Among
these attacks, data integrity attacks can severely impact the efficiency and reliability of edge computing
systems. These integrity issues are mainly because the integrity attack causes inconsistent and unreliable data
that contributes to the distressing edge data analysis process. Therefore, it is crucial to address the data integrity
attack early before it propagates to the other edge nodes. Failure of which can lead to challenging situations to
track the root cause of the attack. All the existing frameworks adopt a security framework to handle data
integrity attacks by including anomaly detections on the existing edge nodes or adding blockchain to identify
the attack. These mechanisms can increase the computational load of edge nodes affecting the lightweight
feature of edge nodes. The proposed work uses a lightweight dimensionality reduction technique to monitor
intrusions and quarantine the data and IoT devices. In future work, this technique will be improved to achieve
better quarantine accuracy.
This research proposes a new concept of data quarantine model to guarantees data integrity in edge
computing. This model can adaptively identify the degree of intervention on the edge data while sourcing the
data from IoT devices. The identified data and devices are alleviated to quarantine model for a predetermined
time without alarming the data analysis process of the neighbouring edge nodes or by adding computational
load. The spam detection module identifies the legitimate devices, and their data is passed to the data scrubber
module. In the scrubber module, sanitized data is passed to the edge layer to continue data analysis. The data
quarantine model shall be implemented and validated with ground truth data for efficiency estimation in future
work.
ACKNOWLEDGEMENT
This research work is supported by Taylor's University, Malaysia, through its Taylor's PhD
Scholarship Program.
REFERENCES
[1] A. M. Alsmadi et al., “Fog computing scheduling algorithm for smart city,” International Journal of Electrical and Computer
Engineering (IJECE), vol. 11, no. 3, pp. 2219–2228, Jun. 2021, doi: 10.11591/ijece.v11i3.pp2219-2228.
[2] M. El Ghmary, Y. Hmimz, T. Chanyour, and M. O. Cherkaoui Malki, “Time and resource constrained offloading with multi-task
in a mobile edge computing node,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 4,
pp. 3757–3766, Aug. 2020, doi: 10.11591/ijece.v10i4.pp3757-3766.
[3] M. Mukherjee et al., “Security and privacy in fog computing: challenges,” IEEE Access, vol. 5, pp. 19293–19304, 2017, doi:
10.1109/ACCESS.2017.2749422.
[4] S. Khan, S. Parkinson, and Y. Qin, “Fog computing security: a review of current applications and security solutions,” Journal of
Cloud Computing, vol. 6, no. 1, Aug. 2017, doi: 10.1186/s13677-017-0090-3.
[5] V. Coskun, E. Cayirci, A. Levi, and S. Sancak, “Quarantine region scheme to mitigate spam attacks in wireless sensor networks,”
IEEE Transactions on Mobile Computing, vol. 5, no. 8, pp. 1074–1086, Aug. 2006, doi: 10.1109/TMC.2006.121.
[6] A. Salis, G. Mancini, R. Bulla, P. Cocco, D. Lezzi, and F. Lordan, “Benefits of a fog-to-cloud approach in proximity marketing,”
in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in
Bioinformatics), vol. 11339 LNCS, Springer International Publishing, 2019, pp. 239–250.
[7] M. Aljarah, M. Shurman, and S. H. Alnabelsi, “Cooperative hierarchical based edge-computing approach for resources allocation
of distributed mobile and IoT applications,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 1,
pp. 296–307, Feb. 2020, doi: 10.11591/ijece.v10i1.pp296-307.

Int J Elec & Comp Eng ISSN: 2088-8708 
A data quarantine model to secure data in edge computing (Poornima Mahadevappa)
3319
[8] G. Lykou, A. Anagnostopoulou, and D. Gritzalis, “Smart airport cybersecurity: threat mitigation and cyber resilience controls,”
Sensors (Switzerland), vol. 19, no. 1, Dec. 2019, Art. no. 19, doi: 10.3390/s19010019.
[9] T. Stremlau, “The financial motivation to keep information secure,” Computer Fraud and Security, vol. 2020, no. 2, pp. 18–19,
2020, doi: 10.1016/S1361-3723(20)30020-8.
[10] B. Cheng, G. Solmaz, F. Cirillo, E. Kovacs, K. Terasawa, and A. Kitazawa, “FogFlow: easy programming of IoT services over
cloud and edges for smart cities,” IEEE Internet of Things Journal, vol. 5, no. 2, pp. 696–707, Apr. 2018, doi:
10.1109/JIOT.2017.2747214.
[11] C. L. Stergiou, K. E. Psannis, and B. B. Gupta, “Iot-based big data secure management in the fog over a 6G wireless network,”
IEEE Internet of Things Journal, vol. 8, no. 7, pp. 5164–5171, Apr. 2021, doi: 10.1109/JIOT.2020.3033131.
[12] J. Zhang, B. Chen, Y. Zhao, X. Cheng, and F. Hu, “Data security and privacy-preserving in edge computing paradigm: survey and
open issues,” IEEE Access, vol. 6, pp. 18209–18237, 2018, doi: 10.1109/ACCESS.2018.2820162.
[13] P. Mahadevappa and R. K. Murugesan, “Study of container-based virtualisation and threats in fog computing,” in Communications
in Computer and Information Science, vol. 1347, Springer Singapore, 2021, pp. 535–549.
[14] R. G. Engoulou, M. Bellaïche, S. Pierre, and A. Quintero, “VANET security surveys,” Computer Communications, vol. 44,
pp. 1–13, May 2014, doi: 10.1016/j.comcom.2014.02.020.
[15] D. Liu, Z. Yan, W. Ding, and M. Atiquzzaman, “A survey on secure data analytics in edge computing,” IEEE Internet of Things
Journal, vol. 6, no. 3, pp. 4946–4967, Jun. 2019, doi: 10.1109/JIOT.2019.2897619.
[16] P. Zhang, J. K. Liu, F. Richard Yu, M. Sookhak, M. H. Au, and X. Luo, “A survey on access control in fog computing,” IEEE
Communications Magazine, vol. 56, no. 2, pp. 144–149, Feb. 2018, doi: 10.1109/MCOM.2018.1700333.
[17] A. Zainab, S. S. Refaat, and O. Bouhali, “Ensemble-based spam detection in smart home IoT devices time series data using machine
learning techniques,” Information (Switzerland), vol. 11, no. 11, p. 344, Jul. 2020, doi: 10.3390/INFO11070344.
[18] M. El Ghmary, T. Chanyour, Y. Hmimz, and M. O. C. Malki, “Efficient multi-task offloading with energy and computational
resources optimization in a mobile edge computing node,” International Journal of Electrical and Computer Engineering (IJECE),
vol. 9, no. 6, pp. 4908–4919, Dec. 2019, doi: 10.11591/ijece.v9i6.pp4908-4919.
[19] S. Tuuli and R. Mika, “Detecting Stuxnet‐like data integrity attacks,” Security and Privacy, vol. 3, no. 5, Mar. 2020, doi:
10.1002/spy2.107.
[20] G. S. Aujla, A. Singh, M. Singh, S. Sharma, N. Kumar, and K. K. R. Choo, “BloCkEd: Blockchain-based secure data processing
framework in edge envisioned V2X environment,” IEEE Transactions on Vehicular Technology, vol. 69, no. 6, pp. 5850–5863,
Jun. 2020, doi: 10.1109/TVT.2020.2972278.
[21] X. Xu, X. Zhang, H. Gao, Y. Xue, L. Qi, and W. Dou, “BeCome: blockchain-enabled computation offloading for IoT in mobile
edge computing,” IEEE Transactions on Industrial Informatics, vol. 16, no. 6, pp. 4187–4195, Jun. 2020, doi:
10.1109/TII.2019.2936869.
[22] Z. Cai, B. Qian, and Y. Xiao, “Edge computing based bad metering data detection,” in 2019 3rd IEEE Conference on Energy
Internet and Energy System Integration: Ubiquitous Energy Network Connecting Everything, EI2 2019, Nov. 2019, pp. 693–698,
doi: 10.1109/EI247390.2019.9062052.
[23] A. Alazeb and B. Panda, “Maintaining data integrity in fog computing based critical infrastructure systems,” in Proceedings - 6th
Annual Conference on Computational Science and Computational Intelligence, CSCI 2019, Dec. 2019, pp. 40–47, doi:
10.1109/CSCI49370.2019.00014.
[24] M. Davoodi, R. Moslemi, W. Song, and J. M. Velni, “A fog-based approach to secure smart grids against data integrity attacks,”
Feb. 2020, doi: 10.1109/ISGT45199.2020.9087790.
[25] I. Lera, C. Guerrero, and C. Juiz, “YAFS: a simulator for IoT scenarios in fog computing,” IEEE Access, vol. 7, pp. 91745–91758,
2019, doi: 10.1109/ACCESS.2019.2927895.
[26] “NSL-KDD | Datasets | Research | Canadian Institute for Cybersecurity | UNB.” Accessed: Apr. 20, 2020. [Online]. Available:
https://www.unb.ca/cic/datasets/nsl.html.
BIOGRAPHIES OF AUTHORS
Poornima Mahadevappa is currently a PhD Scholar in Computer Science at
Taylor's University, Malaysia. She obtained her Master of Engineering in Bioinformatics (2014)
from UVCE, Bangalore University, India, and Bachelor of Engineering in Computer Science
(2008) from Ghousia College of Engineering, Visvesvaraya Technological University, India.
She worked as a lecturer in Pooja Bhagavat Memorial Mahajana Post Graduate Centre and as
an Android developer previously. Her research interest includes edge computing, cyber security,
and data analytics. She can be contacted at email: poornimamahadevappa@sd.taylors.edu.my.
Raja Kumar Murugesan is an Associate Professor of Computer Science, and
Head of Research for the Faculty of Innovation and Technology at Taylor's University,
Malaysia. He has a PhD in Advanced Computer Networks from the Universiti Sains Malaysia.
His research interests include IPv6, Future Internet, Internet Governance, Computer Networks,
Network Security, IoT, Blockchain, and Machine Learning. He is a member of the IEEE and
IEEE Communications Society, Internet Society (ISOC), and associated with the IPv6 Forum,
Asia Pacific Advanced Network Group (APAN), Internet2, and Malaysia Network Operator
Group (MyNOG) member's community. He has held various leadership roles in his academic
career. Raja has given several invited talks and presentations on IPv6, Internet Governance, IoT,
Blockchain, AI, Machine learning and Digital Transformation at various international
conferences and events. He can be contacted at email: rajakumar.murugesan@taylors.edu.my.